[Python] How can I calculate SSIM between 2 images using scikit-image?

I am trying to find the SSIM between two images that I'm storing using the Image data structure from PIL. the structural_similarity function requires both images to be stored as numpy arrays of the same shape, but despite me supplying such arrays, I keep getting an error suggesting they are not in fact the same shape.

The error I keep getting comes from the structural_similarity function, with the error being:

operands could not be broadcast together with shapes (3,) (2,)


My code looks as follows:

import numpy as np
from PIL import Image
from skimage.metrics import structural_similarity

def readAndResize(image_path, width=60, height=30):
try:
image = Image.open(image_path)
image = image.resize((width, height))
return image
except:
print("Error: Could not read the image at", image_path)
return []

def computeMeasure1(image1, image2):
img1 = np.asarray(image1)
img2 = np.asarray(image2)
print(img1.shape) # prints (30, 60, 3)
print(img2.shape) # prints (30, 60, 3)
measure_value = structural_similarity(img1, img2, win_size=(3,3), multichannel=True)
return measure_value

def main():

path = "Images/Student_Test/generated.photos_v3_0228313.jpg"
path2 = "Images/Student_Train/6_img_1553496386570.jpg"

image = (readAndResize(path))
image2 = (readAndResize(path2))

print(computeMeasure1(image,image2))

main()


I have tried to store the images using imread() from opencv-python, but the exact same problem persisted, with the same error popping up despite both arrays being the same shape.

Continue reading...